University of Konstanz
Graduiertenkolleg / PhD Program
Computer and Information Science

Colloquium of the Department and the PhD Program

title

Directed network models of text

speaker

Prof. Steven R. Corman, Hugh Downs School of Human Communication, Arizona State University, USA
Arizona, USA

date & place

Wednesday, 11.07.2007, 16:15 h
Room C252

abstract

The information space defined by text has expanded exponentially with the growth of the Internet, and with it the need for computer-based analysis and visualization. However the primary tools available for this purpose were developed last century and rely on counting methods that reduce texts to "piles of words." Centering Resonance Analysis (CRA, Corman et al., 2002) was developed as a more sophisticated text analysis method. By using linguistic methods to represent texts as networks, it aims to recover relationships between words that were intended by the author of a text. CRA is increasingly applied in academic research. However, because it relies on undirected graphs derived from noun phrases, it does not make full use of the structure embedded in texts, and in some cases creates spurious links between words. This presentation outlines a refinement called Directed Centering Resonance Analysis (DCRA) that applies Case Grammar and Argument Influence Theory to extract mixed graphs from texts in an effort to overcome the limitations of CRA. The technique is described, and then illustrated with application to discourse in the blogosphere about the American presidential race.